Jakub Dotlačil (Groningen)

— joint work with Mojmir Docekal —

In the analysis of negation, it is common to distinguish (at least) three different groups of expressions licensed in negative-like environments: weak negative polarity items (anything, (1a)), strong NPIs (punctual until, (1b)) and n-words. The last group is restricted
to the so-called negative concord (NC) languages, as is Czech, (1c).

(1a) John didn't say anything.
(1b) John didn't leave until his birthday.
(1c) Neznám tady nikoho
Not-know here nobody
I don't know anybody here.

While the licensing conditions of (strong) NPIs are commonly characterized in semantic terms (e.g., Zwarts, 1996, Gajewski, 2011, Chierchia, 2013) and the distribution of n-words is described in syntactic terms (Zeijlstra, 2004, a.o.) it is also often observed that the two groups are closely related (Laka, 1990, a.o.).

In this talk I go through several experiments that study the distribution of negative elements in Czech. We'll see that besides the two uncontroversial groups (weak NPIs, n-words), there is evidence that Czech also has strong NPIs. This has interesting consequences: (i) strong NPIs can be used as a probe of the existence of neg-raising, which has been claimed to be absent in Slavic (Boskovic and Gajewski, 2009), incorrectly, in our view, (ii) since strong NPIs and n-words have almost identical distributions, their co-existence alongside n-words pose a challenge to learning; that is, they raise an issue as to how Czech speakers can correctly categorize some items as strong NPIs and other items as n-words, when there is so little evidence to distinguish the two groups. I will discuss experimental evidence probing issues (i) and (ii).